The Local Retail Discovery Split: Map Pack, Marketplace, and the Rise of the AI Answer

Four layers, one decision

A shopper looking for an independent retailer today rarely makes that decision on a single surface. The choice is split across four layers that behave differently from each other. There is the Google local, or map, pack, the classic three-listing block tied to a physical location. There is organic and marketplace search, the ranked list of links and product listings that has anchored local discovery for two decades. There are near-me and in-stock intent signals, the searches a shopper runs specifically to find something close by and available now. And there is a newer layer that does not list anything at all: an AI answer, whether that is a Google AI Overview, a ChatGPT response, or the conversational search Google has now built directly into Maps, that synthesizes a single response instead of handing the shopper a set of options to compare.

Those four layers are not interchangeable, and they are not moving at the same speed. The map pack and organic results are mature, well-understood surfaces with decades of measurement behind them. The AI-answer layer is new enough that, as the gaps in this study make clear, no independent, methodologically transparent census of it exists yet for local retail specifically. That does not mean the shift is not happening. It means the size of the shift has to be stated carefully, with the tier of evidence behind each number made explicit rather than assumed.

What happens when Google answers instead of lists

Pew Research Center ran the most rigorous test available on this question: a tracked panel of 900 US adults across 68,879 searches, 12,593 of which returned an AI summary. The results were unambiguous. When a summary appeared, users clicked a traditional result link in only 8% of visits, against 15% when no summary appeared, nearly double. Clicking a link inside the summary itself happened in just 1% of visits. And the session was more likely to end there entirely: 26% of AI-summary visits ended with no further click, versus 16% of visits with only traditional results.

By March 2025, this was already a broad phenomenon: about one in five Google searches produced a summary, and 58% of Pew's panel had seen one that month. This is general search behavior, not a local-specific measurement, but it establishes the mechanism that the local-specific numbers later in this study build on. When an answer is synthesized instead of listed, fewer people click through to anything, including your website and your Google Business Profile.

Clicking on a link inside the AI summary itself happened in just 1% of visits.

Local queries and the AI Overview question

The general-search click-suppression data from Pew is established fact. The local-specific version of the same story is not yet at that level of confidence. Search Engine Journal, citing data from the local-SEO vendor Whitespark, reported that Google AI Overviews now appear on 68% of local business-type queries, compared with 39% for the traditional map pack, and that the appearance rate climbs to 92% for informational near-me queries and 97% for queries with hybrid intent. Those are striking numbers, and if they hold up they describe a local search environment where the map pack, the surface local SEO has optimized for since it existed, is no longer the dominant response Google gives.

But the reporting on those figures does not disclose a sample size or geographic scope, which is why this study tiers the claim contested rather than established. That is not a reason to dismiss it. It is a reason to treat it as a strong early signal worth watching in your own market rather than a number to build a strategy on unchecked.

The Google-to-ChatGPT swing: signal or single-vendor noise

The single most dramatic number in this evidence set is also the one that deserves the most scrutiny. BrightLocal's Local Consumer Review Survey 2026, a panel of 1,002 US adults, found that the share of consumers naming Google as their platform for finding local businesses fell from 83% to 71% in one year, while the share naming ChatGPT or other AI tools rose from 6% to 45%, enough to make AI tools the third most-used local-discovery platform in that panel. A separate BrightLocal survey from earlier the same year, also drawn from a single vendor's panel, found 40% of US consumers say they are actively using generative AI within search, and that 10% of Gen Z consumers now consider ChatGPT their default search platform, roughly double the 5% rate across the general population.

These are single-vendor figures with undisclosed methodology behind the year-over-year comparison, which is why this study tiers the headline swing contested. A number moving from 6% to 45% in twelve months in one survey's panel is the kind of result that deserves independent replication before it is treated as settled fact about the whole market. What it is not is noise to be ignored. It is directionally consistent with everything else in this evidence set, from the Pew click-suppression data to the AI Overview appearance rates, and it is exactly the kind of frontier claim that a retailer's own measurement, not a single survey, should be used to confirm or challenge for their specific market.

Reviews are still the trust currency

If fewer clicks are happening and more of the decision gets made inside a synthesized answer, the proof that answer relies on matters more, not less. BrightLocal's review survey found that 97% of consumers read online reviews for local businesses, that 41% say they always do, up sharply from 29% the year before, and that 49% trust online reviews as much as a personal recommendation from someone they know. These figures rest on a single vendor's panel without disclosed methodology, so they are tiered emerging here rather than established, but they line up with the mechanical logic of the AI-answer layer: an engine that synthesizes a response about a local business has to draw its confidence from somewhere, and reviews are one of the most structured, machine-readable trust signals a small retailer controls directly.

About one in five consumers, per that same single-vendor BrightLocal research program, now run local searches directly inside a maps app rather than a general search engine, a second reminder that review volume and quality need to be strong on the specific platforms shoppers are actually opening, not just wherever a retailer happens to have set up a profile first.

Google is building the interface directly into maps

The clearest piece of established, verifiable fact in this study about where the AI-answer layer is heading is not a survey at all. It is a product launch. Google shipped Ask Maps, a Gemini-powered conversational search feature built directly into Google Maps, first in the US and India, on Android and iOS, with desktop support described as coming later. Google says the feature draws on a database of more than 300 million places and reviews from 500 million contributors, and no advertising inside the feature was disclosed at launch.

This matters for two reasons beyond the headline. First, it is the clearest evidence that the AI-answer layer is not confined to the general search results page; it is being built into the app shoppers already use specifically to find something nearby. Second, it is the one genuinely global data point in this evidence set. Every other figure here comes from a US panel or US federal data. Ask Maps launching in India alongside the US is a signal that this shift in how local discovery works is not a US-only story, even if the behavioral measurement of it currently is.

Ask Maps launched first in the US and India together, on a database Google describes as covering more than 300 million places.

Can visibility inside the answer be engineered

A 2023 preprint on generative-engine optimization, produced by researchers including collaborators from Princeton, found that content-level optimization strategies could raise a source's visibility inside generative-engine responses by up to 40%. That is a meaningful number, but two caveats belong with it. It is a single non-peer-reviewed preprint whose headline figure has not been independently replicated, so it is tiered emerging here rather than established. And the paper's own finding is that effectiveness varies substantially by content domain, meaning no single tactic generalizes across categories, which is a genuinely useful piece of humility for anyone selling a one-size-fits-all playbook for getting cited by an AI engine.

What this paper does establish, even at emerging tier, is that visibility inside a synthesized answer is not purely random and not purely a function of brand size. It responds to how content is structured. That is consistent with everything else in this study: reviews, structured data, and profile completeness are not just SEO housekeeping anymore, they are the raw material an AI engine draws from when it decides which business to name.

The physical store still wins the transaction, but not the decision

It is worth returning, at the end of this study, to the number that grounds all of it: 16.9% of US retail sales happened online in the first quarter of 2026, per the Census Bureau, meaning the large majority of retail dollars, roughly 83%, still transacted through a physical location. The question this study set out to answer was never whether independent retail would move online. It has not, at least not yet, at anything close to the scale the AI-search headlines might suggest. The question was what decides which physical door a shopper walks through: increasingly, it runs through a search that happens before they ever leave home.

Aggregated figures from Backlinko, tiered contested here because several are undated and attributed only generically to Google in current reporting, describe a local-intent funnel with real urgency behind it: 88% of smartphone users who run a local search visit a related store within a week, and 76% of near-me searchers visit within a day. Whatever the exact percentages turn out to be under closer independent scrutiny, the shape of the story holds across every tier of evidence in this study. The transaction is still overwhelmingly physical. The decision about where that transaction happens is increasingly made somewhere an independent retailer cannot see, inside an AI answer, before a human ever sees a list.